1 /**************************************************************************
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
30 #include "main/glheader.h"
31 #include "main/bufferobj.h"
32 #include "main/context.h"
33 #include "main/enums.h"
36 #include "brw_defines.h"
37 #include "brw_context.h"
38 #include "brw_state.h"
40 #include "intel_batchbuffer.h"
41 #include "intel_buffer_objects.h"
43 static GLuint double_types
[5] = {
45 BRW_SURFACEFORMAT_R64_FLOAT
,
46 BRW_SURFACEFORMAT_R64G64_FLOAT
,
47 BRW_SURFACEFORMAT_R64G64B64_FLOAT
,
48 BRW_SURFACEFORMAT_R64G64B64A64_FLOAT
51 static GLuint float_types
[5] = {
53 BRW_SURFACEFORMAT_R32_FLOAT
,
54 BRW_SURFACEFORMAT_R32G32_FLOAT
,
55 BRW_SURFACEFORMAT_R32G32B32_FLOAT
,
56 BRW_SURFACEFORMAT_R32G32B32A32_FLOAT
59 static GLuint half_float_types
[5] = {
61 BRW_SURFACEFORMAT_R16_FLOAT
,
62 BRW_SURFACEFORMAT_R16G16_FLOAT
,
63 BRW_SURFACEFORMAT_R16G16B16A16_FLOAT
,
64 BRW_SURFACEFORMAT_R16G16B16A16_FLOAT
67 static GLuint uint_types_norm
[5] = {
69 BRW_SURFACEFORMAT_R32_UNORM
,
70 BRW_SURFACEFORMAT_R32G32_UNORM
,
71 BRW_SURFACEFORMAT_R32G32B32_UNORM
,
72 BRW_SURFACEFORMAT_R32G32B32A32_UNORM
75 static GLuint uint_types_scale
[5] = {
77 BRW_SURFACEFORMAT_R32_USCALED
,
78 BRW_SURFACEFORMAT_R32G32_USCALED
,
79 BRW_SURFACEFORMAT_R32G32B32_USCALED
,
80 BRW_SURFACEFORMAT_R32G32B32A32_USCALED
83 static GLuint int_types_norm
[5] = {
85 BRW_SURFACEFORMAT_R32_SNORM
,
86 BRW_SURFACEFORMAT_R32G32_SNORM
,
87 BRW_SURFACEFORMAT_R32G32B32_SNORM
,
88 BRW_SURFACEFORMAT_R32G32B32A32_SNORM
91 static GLuint int_types_scale
[5] = {
93 BRW_SURFACEFORMAT_R32_SSCALED
,
94 BRW_SURFACEFORMAT_R32G32_SSCALED
,
95 BRW_SURFACEFORMAT_R32G32B32_SSCALED
,
96 BRW_SURFACEFORMAT_R32G32B32A32_SSCALED
99 static GLuint ushort_types_norm
[5] = {
101 BRW_SURFACEFORMAT_R16_UNORM
,
102 BRW_SURFACEFORMAT_R16G16_UNORM
,
103 BRW_SURFACEFORMAT_R16G16B16_UNORM
,
104 BRW_SURFACEFORMAT_R16G16B16A16_UNORM
107 static GLuint ushort_types_scale
[5] = {
109 BRW_SURFACEFORMAT_R16_USCALED
,
110 BRW_SURFACEFORMAT_R16G16_USCALED
,
111 BRW_SURFACEFORMAT_R16G16B16_USCALED
,
112 BRW_SURFACEFORMAT_R16G16B16A16_USCALED
115 static GLuint short_types_norm
[5] = {
117 BRW_SURFACEFORMAT_R16_SNORM
,
118 BRW_SURFACEFORMAT_R16G16_SNORM
,
119 BRW_SURFACEFORMAT_R16G16B16_SNORM
,
120 BRW_SURFACEFORMAT_R16G16B16A16_SNORM
123 static GLuint short_types_scale
[5] = {
125 BRW_SURFACEFORMAT_R16_SSCALED
,
126 BRW_SURFACEFORMAT_R16G16_SSCALED
,
127 BRW_SURFACEFORMAT_R16G16B16_SSCALED
,
128 BRW_SURFACEFORMAT_R16G16B16A16_SSCALED
131 static GLuint ubyte_types_norm
[5] = {
133 BRW_SURFACEFORMAT_R8_UNORM
,
134 BRW_SURFACEFORMAT_R8G8_UNORM
,
135 BRW_SURFACEFORMAT_R8G8B8_UNORM
,
136 BRW_SURFACEFORMAT_R8G8B8A8_UNORM
139 static GLuint ubyte_types_scale
[5] = {
141 BRW_SURFACEFORMAT_R8_USCALED
,
142 BRW_SURFACEFORMAT_R8G8_USCALED
,
143 BRW_SURFACEFORMAT_R8G8B8_USCALED
,
144 BRW_SURFACEFORMAT_R8G8B8A8_USCALED
147 static GLuint byte_types_norm
[5] = {
149 BRW_SURFACEFORMAT_R8_SNORM
,
150 BRW_SURFACEFORMAT_R8G8_SNORM
,
151 BRW_SURFACEFORMAT_R8G8B8_SNORM
,
152 BRW_SURFACEFORMAT_R8G8B8A8_SNORM
155 static GLuint byte_types_scale
[5] = {
157 BRW_SURFACEFORMAT_R8_SSCALED
,
158 BRW_SURFACEFORMAT_R8G8_SSCALED
,
159 BRW_SURFACEFORMAT_R8G8B8_SSCALED
,
160 BRW_SURFACEFORMAT_R8G8B8A8_SSCALED
165 * Given vertex array type/size/format/normalized info, return
166 * the appopriate hardware surface type.
167 * Format will be GL_RGBA or possibly GL_BGRA for GLubyte[4] color arrays.
169 static GLuint
get_surface_type( GLenum type
, GLuint size
,
170 GLenum format
, GLboolean normalized
)
172 if (unlikely(INTEL_DEBUG
& DEBUG_VERTS
))
173 printf("type %s size %d normalized %d\n",
174 _mesa_lookup_enum_by_nr(type
), size
, normalized
);
178 case GL_DOUBLE
: return double_types
[size
];
179 case GL_FLOAT
: return float_types
[size
];
180 case GL_HALF_FLOAT
: return half_float_types
[size
];
181 case GL_INT
: return int_types_norm
[size
];
182 case GL_SHORT
: return short_types_norm
[size
];
183 case GL_BYTE
: return byte_types_norm
[size
];
184 case GL_UNSIGNED_INT
: return uint_types_norm
[size
];
185 case GL_UNSIGNED_SHORT
: return ushort_types_norm
[size
];
186 case GL_UNSIGNED_BYTE
:
187 if (format
== GL_BGRA
) {
188 /* See GL_EXT_vertex_array_bgra */
190 return BRW_SURFACEFORMAT_B8G8R8A8_UNORM
;
193 return ubyte_types_norm
[size
];
195 default: assert(0); return 0;
199 assert(format
== GL_RGBA
); /* sanity check */
201 case GL_DOUBLE
: return double_types
[size
];
202 case GL_FLOAT
: return float_types
[size
];
203 case GL_HALF_FLOAT
: return half_float_types
[size
];
204 case GL_INT
: return int_types_scale
[size
];
205 case GL_SHORT
: return short_types_scale
[size
];
206 case GL_BYTE
: return byte_types_scale
[size
];
207 case GL_UNSIGNED_INT
: return uint_types_scale
[size
];
208 case GL_UNSIGNED_SHORT
: return ushort_types_scale
[size
];
209 case GL_UNSIGNED_BYTE
: return ubyte_types_scale
[size
];
210 default: assert(0); return 0;
216 static GLuint
get_size( GLenum type
)
219 case GL_DOUBLE
: return sizeof(GLdouble
);
220 case GL_FLOAT
: return sizeof(GLfloat
);
221 case GL_HALF_FLOAT
: return sizeof(GLhalfARB
);
222 case GL_INT
: return sizeof(GLint
);
223 case GL_SHORT
: return sizeof(GLshort
);
224 case GL_BYTE
: return sizeof(GLbyte
);
225 case GL_UNSIGNED_INT
: return sizeof(GLuint
);
226 case GL_UNSIGNED_SHORT
: return sizeof(GLushort
);
227 case GL_UNSIGNED_BYTE
: return sizeof(GLubyte
);
232 static GLuint
get_index_type(GLenum type
)
235 case GL_UNSIGNED_BYTE
: return BRW_INDEX_BYTE
;
236 case GL_UNSIGNED_SHORT
: return BRW_INDEX_WORD
;
237 case GL_UNSIGNED_INT
: return BRW_INDEX_DWORD
;
238 default: assert(0); return 0;
243 copy_array_to_vbo_array( struct brw_context
*brw
,
244 struct brw_vertex_element
*element
,
245 struct brw_vertex_buffer
*buffer
,
248 GLuint size
= element
->count
* dst_stride
;
250 buffer
->stride
= dst_stride
;
251 if (dst_stride
== element
->glarray
->StrideB
) {
252 intel_upload_data(&brw
->intel
, element
->glarray
->Ptr
, size
, dst_stride
,
253 &buffer
->bo
, &buffer
->offset
);
255 const unsigned char *src
= element
->glarray
->Ptr
;
256 char *dst
= intel_upload_map(&brw
->intel
, size
, dst_stride
);
259 for (i
= 0; i
< element
->count
; i
++) {
260 memcpy(dst
, src
, dst_stride
);
261 src
+= element
->glarray
->StrideB
;
264 intel_upload_unmap(&brw
->intel
, dst
, size
, dst_stride
,
265 &buffer
->bo
, &buffer
->offset
);
269 static void brw_prepare_vertices(struct brw_context
*brw
)
271 struct gl_context
*ctx
= &brw
->intel
.ctx
;
272 struct intel_context
*intel
= intel_context(ctx
);
273 GLbitfield vs_inputs
= brw
->vs
.prog_data
->inputs_read
;
274 const unsigned char *ptr
= NULL
;
275 GLuint interleaved
= 0, total_size
= 0;
276 unsigned int min_index
= brw
->vb
.min_index
;
277 unsigned int max_index
= brw
->vb
.max_index
;
280 struct brw_vertex_element
*upload
[VERT_ATTRIB_MAX
];
281 GLuint nr_uploads
= 0;
283 /* First build an array of pointers to ve's in vb.inputs_read
286 printf("%s %d..%d\n", __FUNCTION__
, min_index
, max_index
);
288 /* Accumulate the list of enabled arrays. */
289 brw
->vb
.nr_enabled
= 0;
291 GLuint i
= _mesa_ffsll(vs_inputs
) - 1;
292 struct brw_vertex_element
*input
= &brw
->vb
.inputs
[i
];
294 vs_inputs
&= ~(1 << i
);
295 brw
->vb
.enabled
[brw
->vb
.nr_enabled
++] = input
;
298 if (brw
->vb
.nr_enabled
== 0)
301 if (brw
->vb
.nr_buffers
)
304 /* XXX: In the rare cases where this happens we fallback all
305 * the way to software rasterization, although a tnl fallback
306 * would be sufficient. I don't know of *any* real world
307 * cases with > 17 vertex attributes enabled, so it probably
308 * isn't an issue at this point.
310 if (brw
->vb
.nr_enabled
>= BRW_VEP_MAX
) {
311 intel
->Fallback
= GL_TRUE
; /* boolean, not bitfield */
315 for (i
= j
= 0; i
< brw
->vb
.nr_enabled
; i
++) {
316 struct brw_vertex_element
*input
= brw
->vb
.enabled
[i
];
317 const struct gl_client_array
*glarray
= input
->glarray
;
318 int type_size
= get_size(glarray
->Type
);
320 input
->element_size
= type_size
* glarray
->Size
;
322 if (_mesa_is_bufferobj(glarray
->BufferObj
)) {
323 struct intel_buffer_object
*intel_buffer
=
324 intel_buffer_object(glarray
->BufferObj
);
327 for (k
= 0; k
< i
; k
++) {
328 const struct gl_client_array
*other
= brw
->vb
.enabled
[k
]->glarray
;
329 if (glarray
->BufferObj
== other
->BufferObj
&&
330 glarray
->StrideB
== other
->StrideB
&&
331 (uintptr_t)(glarray
->Ptr
- other
->Ptr
) < glarray
->StrideB
)
333 input
->buffer
= brw
->vb
.enabled
[k
]->buffer
;
334 input
->offset
= glarray
->Ptr
- other
->Ptr
;
339 struct brw_vertex_buffer
*buffer
= &brw
->vb
.buffers
[j
];
340 /* Named buffer object: Just reference its contents directly. */
341 buffer
->bo
= intel_bufferobj_source(intel
, intel_buffer
,
343 drm_intel_bo_reference(buffer
->bo
);
344 buffer
->offset
+= (uintptr_t)glarray
->Ptr
;
345 buffer
->stride
= glarray
->StrideB
;
350 input
->count
= glarray
->_MaxElement
;
352 /* This is a common place to reach if the user mistakenly supplies
353 * a pointer in place of a VBO offset. If we just let it go through,
354 * we may end up dereferencing a pointer beyond the bounds of the
355 * GTT. We would hope that the VBO's max_index would save us, but
356 * Mesa appears to hand us min/max values not clipped to the
357 * array object's _MaxElement, and _MaxElement frequently appears
358 * to be wrong anyway.
360 * The VBO spec allows application termination in this case, and it's
361 * probably a service to the poor programmer to do so rather than
362 * trying to just not render.
364 assert(input
->offset
< brw
->vb
.buffers
[input
->buffer
].bo
->size
);
366 input
->count
= glarray
->StrideB
? max_index
+ 1 : 1;
368 /* Queue the buffer object up to be uploaded in the next pass,
369 * when we've decided if we're doing interleaved or not.
371 if (nr_uploads
== 0) {
372 /* Position array not properly enabled:
374 if (input
->attrib
== VERT_ATTRIB_POS
&& glarray
->StrideB
== 0) {
375 intel
->Fallback
= GL_TRUE
; /* boolean, not bitfield */
379 interleaved
= glarray
->StrideB
;
382 else if (interleaved
!= glarray
->StrideB
||
383 (uintptr_t)(glarray
->Ptr
- ptr
) > interleaved
)
387 else if (total_size
& (type_size
-1))
389 /* enforce natural alignment (for doubles) */
393 upload
[nr_uploads
++] = input
;
394 total_size
+= input
->element_size
;
398 /* Handle any arrays to be uploaded. */
399 if (nr_uploads
> 1) {
400 if (interleaved
&& interleaved
<= 2*total_size
) {
401 /* All uploads are interleaved, so upload the arrays together as
402 * interleaved. First, upload the contents and set up upload[0].
404 copy_array_to_vbo_array(brw
,
405 upload
[0], &brw
->vb
.buffers
[j
],
408 for (i
= 0; i
< nr_uploads
; i
++) {
409 /* Then, just point upload[i] at upload[0]'s buffer. */
411 ((const unsigned char *)upload
[i
]->glarray
->Ptr
- ptr
);
412 upload
[i
]->buffer
= j
;
418 else if (total_size
< 2048) {
419 /* Upload non-interleaved arrays into a single interleaved array */
420 struct brw_vertex_buffer
*buffer
= &brw
->vb
.buffers
[j
];
421 int count
= upload
[0]->count
, offset
;
424 map
= intel_upload_map(&brw
->intel
, total_size
* count
, total_size
);
425 for (i
= offset
= 0; i
< nr_uploads
; i
++) {
426 const unsigned char *src
= upload
[i
]->glarray
->Ptr
;
427 int size
= upload
[i
]->element_size
;
428 int stride
= upload
[i
]->glarray
->StrideB
;
429 char *dst
= map
+ offset
;
432 for (n
= 0; n
< count
; n
++) {
433 memcpy(dst
, src
, size
);
438 upload
[i
]->offset
= offset
;
439 upload
[i
]->buffer
= j
;
443 intel_upload_unmap(&brw
->intel
, map
, total_size
* count
, total_size
,
444 &buffer
->bo
, &buffer
->offset
);
445 buffer
->stride
= offset
;
451 /* Upload non-interleaved arrays */
452 for (i
= 0; i
< nr_uploads
; i
++) {
453 copy_array_to_vbo_array(brw
,
454 upload
[i
], &brw
->vb
.buffers
[j
],
455 upload
[i
]->element_size
);
456 upload
[i
]->buffer
= j
++;
459 /* can we simply extend the current vb? */
460 brw
->vb
.start_vertex_bias
= 0;
461 if (j
== brw
->vb
.nr_current_buffers
) {
463 for (i
= 0; i
< j
; i
++) {
466 if (brw
->vb
.current_buffers
[i
].handle
!= brw
->vb
.buffers
[i
].bo
->handle
||
467 brw
->vb
.current_buffers
[i
].stride
!= brw
->vb
.buffers
[i
].stride
)
470 d
= brw
->vb
.buffers
[i
].offset
- brw
->vb
.current_buffers
[i
].offset
;
472 delta
= d
/ brw
->vb
.current_buffers
[i
].stride
;
473 else if (delta
* brw
->vb
.current_buffers
[i
].stride
!= d
)
478 brw
->vb
.start_vertex_bias
= delta
;
480 drm_intel_bo_unreference(brw
->vb
.buffers
[j
].bo
);
485 brw
->vb
.nr_buffers
= j
;
488 brw_prepare_query_begin(brw
);
489 for (i
= 0; i
< brw
->vb
.nr_buffers
; i
++) {
490 brw_add_validated_bo(brw
, brw
->vb
.buffers
[i
].bo
);
494 static void brw_emit_vertices(struct brw_context
*brw
)
496 struct gl_context
*ctx
= &brw
->intel
.ctx
;
497 struct intel_context
*intel
= intel_context(ctx
);
500 brw_emit_query_begin(brw
);
502 /* If the VS doesn't read any inputs (calculating vertex position from
503 * a state variable for some reason, for example), emit a single pad
504 * VERTEX_ELEMENT struct and bail.
506 * The stale VB state stays in place, but they don't do anything unless
507 * a VE loads from them.
509 if (brw
->vb
.nr_enabled
== 0) {
511 OUT_BATCH((CMD_VERTEX_ELEMENT
<< 16) | 1);
512 if (intel
->gen
>= 6) {
513 OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT
) |
515 (BRW_SURFACEFORMAT_R32G32B32A32_FLOAT
<< BRW_VE0_FORMAT_SHIFT
) |
516 (0 << BRW_VE0_SRC_OFFSET_SHIFT
));
518 OUT_BATCH((0 << BRW_VE0_INDEX_SHIFT
) |
520 (BRW_SURFACEFORMAT_R32G32B32A32_FLOAT
<< BRW_VE0_FORMAT_SHIFT
) |
521 (0 << BRW_VE0_SRC_OFFSET_SHIFT
));
523 OUT_BATCH((BRW_VE1_COMPONENT_STORE_0
<< BRW_VE1_COMPONENT_0_SHIFT
) |
524 (BRW_VE1_COMPONENT_STORE_0
<< BRW_VE1_COMPONENT_1_SHIFT
) |
525 (BRW_VE1_COMPONENT_STORE_0
<< BRW_VE1_COMPONENT_2_SHIFT
) |
526 (BRW_VE1_COMPONENT_STORE_1_FLT
<< BRW_VE1_COMPONENT_3_SHIFT
));
531 /* Now emit VB and VEP state packets.
534 if (brw
->vb
.nr_buffers
) {
535 BEGIN_BATCH(1 + 4*brw
->vb
.nr_buffers
);
536 OUT_BATCH((CMD_VERTEX_BUFFER
<< 16) | (4*brw
->vb
.nr_buffers
- 1));
537 for (i
= 0; i
< brw
->vb
.nr_buffers
; i
++) {
538 struct brw_vertex_buffer
*buffer
= &brw
->vb
.buffers
[i
];
541 if (intel
->gen
>= 6) {
542 dw0
= GEN6_VB0_ACCESS_VERTEXDATA
| (i
<< GEN6_VB0_INDEX_SHIFT
);
544 dw0
= BRW_VB0_ACCESS_VERTEXDATA
| (i
<< BRW_VB0_INDEX_SHIFT
);
547 OUT_BATCH(dw0
| (buffer
->stride
<< BRW_VB0_PITCH_SHIFT
));
548 OUT_RELOC(buffer
->bo
, I915_GEM_DOMAIN_VERTEX
, 0, buffer
->offset
);
549 if (intel
->gen
>= 5) {
550 OUT_RELOC(buffer
->bo
, I915_GEM_DOMAIN_VERTEX
, 0, buffer
->bo
->size
- 1);
553 OUT_BATCH(0); /* Instance data step rate */
555 brw
->vb
.current_buffers
[i
].handle
= buffer
->bo
->handle
;
556 brw
->vb
.current_buffers
[i
].offset
= buffer
->offset
;
557 brw
->vb
.current_buffers
[i
].stride
= buffer
->stride
;
559 brw
->vb
.nr_current_buffers
= i
;
563 BEGIN_BATCH(1 + brw
->vb
.nr_enabled
* 2);
564 OUT_BATCH((CMD_VERTEX_ELEMENT
<< 16) | (2*brw
->vb
.nr_enabled
- 1));
565 for (i
= 0; i
< brw
->vb
.nr_enabled
; i
++) {
566 struct brw_vertex_element
*input
= brw
->vb
.enabled
[i
];
567 uint32_t format
= get_surface_type(input
->glarray
->Type
,
568 input
->glarray
->Size
,
569 input
->glarray
->Format
,
570 input
->glarray
->Normalized
);
571 uint32_t comp0
= BRW_VE1_COMPONENT_STORE_SRC
;
572 uint32_t comp1
= BRW_VE1_COMPONENT_STORE_SRC
;
573 uint32_t comp2
= BRW_VE1_COMPONENT_STORE_SRC
;
574 uint32_t comp3
= BRW_VE1_COMPONENT_STORE_SRC
;
576 switch (input
->glarray
->Size
) {
577 case 0: comp0
= BRW_VE1_COMPONENT_STORE_0
;
578 case 1: comp1
= BRW_VE1_COMPONENT_STORE_0
;
579 case 2: comp2
= BRW_VE1_COMPONENT_STORE_0
;
580 case 3: comp3
= BRW_VE1_COMPONENT_STORE_1_FLT
;
584 if (intel
->gen
>= 6) {
585 OUT_BATCH((input
->buffer
<< GEN6_VE0_INDEX_SHIFT
) |
587 (format
<< BRW_VE0_FORMAT_SHIFT
) |
588 (input
->offset
<< BRW_VE0_SRC_OFFSET_SHIFT
));
590 OUT_BATCH((input
->buffer
<< BRW_VE0_INDEX_SHIFT
) |
592 (format
<< BRW_VE0_FORMAT_SHIFT
) |
593 (input
->offset
<< BRW_VE0_SRC_OFFSET_SHIFT
));
597 OUT_BATCH((comp0
<< BRW_VE1_COMPONENT_0_SHIFT
) |
598 (comp1
<< BRW_VE1_COMPONENT_1_SHIFT
) |
599 (comp2
<< BRW_VE1_COMPONENT_2_SHIFT
) |
600 (comp3
<< BRW_VE1_COMPONENT_3_SHIFT
));
602 OUT_BATCH((comp0
<< BRW_VE1_COMPONENT_0_SHIFT
) |
603 (comp1
<< BRW_VE1_COMPONENT_1_SHIFT
) |
604 (comp2
<< BRW_VE1_COMPONENT_2_SHIFT
) |
605 (comp3
<< BRW_VE1_COMPONENT_3_SHIFT
) |
606 ((i
* 4) << BRW_VE1_DST_OFFSET_SHIFT
));
611 const struct brw_tracked_state brw_vertices
= {
614 .brw
= BRW_NEW_BATCH
| BRW_NEW_VERTICES
,
617 .prepare
= brw_prepare_vertices
,
618 .emit
= brw_emit_vertices
,
621 static void brw_prepare_indices(struct brw_context
*brw
)
623 struct gl_context
*ctx
= &brw
->intel
.ctx
;
624 struct intel_context
*intel
= &brw
->intel
;
625 const struct _mesa_index_buffer
*index_buffer
= brw
->ib
.ib
;
627 drm_intel_bo
*bo
= NULL
;
628 struct gl_buffer_object
*bufferobj
;
632 if (index_buffer
== NULL
)
635 ib_type_size
= get_size(index_buffer
->type
);
636 ib_size
= ib_type_size
* index_buffer
->count
;
637 bufferobj
= index_buffer
->obj
;
639 /* Turn into a proper VBO:
641 if (!_mesa_is_bufferobj(bufferobj
)) {
643 /* Get new bufferobj, offset:
645 intel_upload_data(&brw
->intel
, index_buffer
->ptr
, ib_size
, ib_type_size
,
647 brw
->ib
.start_vertex_offset
= offset
/ ib_type_size
;
650 offset
= (GLuint
) (unsigned long) index_buffer
->ptr
;
652 /* If the index buffer isn't aligned to its element size, we have to
653 * rebase it into a temporary.
655 if ((get_size(index_buffer
->type
) - 1) & offset
) {
656 GLubyte
*map
= ctx
->Driver
.MapBuffer(ctx
,
657 GL_ELEMENT_ARRAY_BUFFER_ARB
,
662 intel_upload_data(&brw
->intel
, map
, ib_size
, ib_type_size
,
664 brw
->ib
.start_vertex_offset
= offset
/ ib_type_size
;
667 ctx
->Driver
.UnmapBuffer(ctx
, GL_ELEMENT_ARRAY_BUFFER_ARB
, bufferobj
);
669 /* Use CMD_3D_PRIM's start_vertex_offset to avoid re-uploading
670 * the index buffer state when we're just moving the start index
673 brw
->ib
.start_vertex_offset
= offset
/ ib_type_size
;
675 bo
= intel_bufferobj_source(intel
, intel_buffer_object(bufferobj
),
677 drm_intel_bo_reference(bo
);
681 if (brw
->ib
.bo
!= bo
|| brw
->ib
.offset
!= offset
) {
682 drm_intel_bo_unreference(brw
->ib
.bo
);
684 brw
->ib
.offset
= offset
;
686 brw_add_validated_bo(brw
, brw
->ib
.bo
);
687 brw
->state
.dirty
.brw
|= BRW_NEW_INDEX_BUFFER
;
689 drm_intel_bo_unreference(bo
);
693 const struct brw_tracked_state brw_indices
= {
696 .brw
= BRW_NEW_INDICES
,
699 .prepare
= brw_prepare_indices
,
702 static void brw_emit_index_buffer(struct brw_context
*brw
)
704 struct intel_context
*intel
= &brw
->intel
;
705 const struct _mesa_index_buffer
*index_buffer
= brw
->ib
.ib
;
707 if (index_buffer
== NULL
)
711 OUT_BATCH(CMD_INDEX_BUFFER
<< 16 |
712 /* cut index enable << 10 */
713 get_index_type(index_buffer
->type
) << 8 |
715 OUT_RELOC(brw
->ib
.bo
,
716 I915_GEM_DOMAIN_VERTEX
, 0,
718 OUT_RELOC(brw
->ib
.bo
,
719 I915_GEM_DOMAIN_VERTEX
, 0,
720 brw
->ib
.bo
->size
- 1);
724 const struct brw_tracked_state brw_index_buffer
= {
727 .brw
= BRW_NEW_BATCH
| BRW_NEW_INDEX_BUFFER
,
730 .emit
= brw_emit_index_buffer
,